cosmos-rl

Quick Start

  • Installation
  • Single node example
  • Configuration
  • Dataset & Process
  • Customization
  • Hugging Face Model Support

Rollout

  • vLLM
  • [Experimental] TensorRT-LLM

Multi nodes training

  • Multi-node example
  • DGXC-Lepton Job
  • Slurm Job

Elastic & Fault Tolerance

  • Elastic Scaling and Fault Tolerance

Async RL

  • Overview
  • Key Features
  • Architecture
  • Putting It All Together

Parallelism

  • Parallelism

Quantization

  • FP8 Quantization
cosmos-rl
  • Search

© Copyright 2025, NVIDIA.

Built with Sphinx using a theme provided by Read the Docs.